Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userPattern-matching Unit for Medical Applications
73
0
0.0
(
0
)
Authors
O.Leombruni
Ask ChatGPT about the research
No Arabic abstract
We explore the application of concepts developed in High Energy Physics (HEP) for advanced medical data analysis. Our study case is a problem with high social impact: clinically-feasible Magnetic Resonance Fingerprinting (MRF). MRF is a new, quantitative, imaging technique that replaces multiple qualitative Magnetic Resonance Imaging (MRI) exams with a single, reproducible measurement for increased sensitivity and efficiency. A fast acquisition is followed by a pattern matching (PM) task, where signal responses are matched to entries from a dictionary of simulated, physically-feasible responses, yielding multiple tissue parameters simultaneously. Each pixel signal response in the volume is compared through scalar products with all dictionary entries to choose the best measurement reproduction. MRF is limited by the PM processing time, which scales exponentially with the dictionary dimensionality, i.e. with the number of tissue parameters to be reconstructed. We developed for HEP a powerful, compact, embedded system, optimized for extremely fast PM. This system executes real-time tracking for online event selection in the HEP experiments, exploiting maximum parallelism and pipelining. Track reconstruction is executed in two steps. The Associative Memory (AM) ASIC first implements a PM algorithm by recognizing track candidates at low resolution. The second step, which is implemented into FPGAs (Field Programmable Gate Arrays), refines the AM output finding the track parameters at full resolution. We propose to use this system to perform MRF, to achieve clinically reasonable reconstruction time. This paper proposes an adaptation of the HEP system for medical imaging and shows some preliminary results.
rate research
Read More
Brain Imaging Data Structure (BIDS) allows the user to organise brain imaging data into a clear and easy standard directory structure. BIDS is widely supported by the scientific community and is considered a powerful standard for management. The original BIDS is limited to images or data related to the brain. Medical Imaging Data Structure (MIDS) was therefore conceived with the objective of extending this methodology to other anatomical regions and other types of imaging systems in these areas.
Artificial intelligence (AI)-based methods are showing promise in multiple medical-imaging applications. Thus, there is substantial interest in clinical translation of these methods, requiring in turn, that they be evaluated rigorously. In this paper, our goal is to lay out a framework for objective task-based evaluation of AI methods. We will also provide a list of tools available in the literature to conduct this evaluation. Further, we outline the important role of physicians in conducting these evaluation studies. The examples in this paper will be proposed in the context of PET with a focus on neural-network-based methods. However, the framework is also applicable to evaluate other medical-imaging modalities and other types of AI methods.
Since the advent of deep convolutional neural networks (DNNs), computer vision has seen an extremely rapid progress that has led to huge advances in medical imaging. This article does not aim to cover all aspects of the field but focuses on a particular topic, image-to-image translation. Although the topic may not sound familiar, it turns out that many seemingly irrelevant applications can be understood as instances of image-to-image translation. Such applications include (1) noise reduction, (2) super-resolution, (3) image synthesis, and (4) reconstruction. The same underlying principles and algorithms work for various tasks. Our aim is to introduce some of the key ideas on this topic from a uniform point of view. We introduce core ideas and jargon that are specific to image processing by use of DNNs. Having an intuitive grasp of the core ideas of and a knowledge of technical terms would be of great help to the reader for understanding the existing and future applications. Most of the recent applications which build on image-to-image translation are based on one of two fundamental architectures, called pix2pix and CycleGAN, depending on whether the available training data are paired or unpaired. We provide computer codes which implement these two architectures with various enhancements. Our codes are available online with use of the very permissive MIT license. We provide a hands-on tutorial for training a model for denoising based on our codes. We hope that this article, together with the codes, will provide both an overview and the details of the key algorithms, and that it will serve as a basis for the development of new applications.
We consider several types of internal queries: questions about subwords of a text. As the main tool we develop an optimal data structure for the problem called here internal pattern matching. This data structure provides constant-time answers to queries about occurrences of one subword $x$ in another subword $y$ of a given text, assuming that $|y|=mathcal{O}(|x|)$, which allows for a constant-space representation of all occurrences. This problem can be viewed as a natural extension of the well-studied pattern matching problem. The data structure has linear size and admits a linear-time construction algorithm. Using the solution to the internal pattern matching problem, we obtain very efficient data structures answering queries about: primitivity of subwords, periods of subwords, general substring compression, and cyclic equivalence of two subwords. All these results improve upon the best previously known counterparts. The linear construction time of our data structure also allows to improve the algorithm for finding $delta$-subrepetitions in a text (a more general version of maximal repetitions, also called runs). For any fixed $delta$ we obtain the first linear-time algorithm, which matches the linear time complexity of the algorithm computing runs. Our data structure has already been used as a part of the efficient solutions for subword suffix rank & selection, as well as substring compression using Burrows-Wheeler transform composed with run-length encoding.
Deep learning has become the mainstream technology in computer vision, and it has received extensive research interest in developing new medical image processing algorithms to support disease detection and diagnosis. As compared to conventional machine learning technologies, the major advantage of deep learning is that models can automatically identify and recognize representative features through the hierarchal model architecture, while avoiding the laborious development of hand-crafted features. In this paper, we reviewed and summarized more than 200 recently published papers to provide a comprehensive overview of applying deep learning methods in various medical image analysis tasks. Especially, we emphasize the latest progress and contributions of state-of-the-art unsupervised and semi-supervised deep learning in medical images, which are summarized based on different application scenarios, including lesion classification, segmentation, detection, and image registration. Additionally, we also discussed the major technical challenges and suggested the possible solutions in future research efforts.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
